Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Regarding inverted mesh elements

Please login with a confirmed email address before reporting spam

Hi all,
We were doing simulation on eddy current damping pendulum,.While solving the equations ,we got the error message that inverted mesh elements were found. How to avoid this?? can anyone tell us how to solve the equation system for atleast a compound pendulum??

Kindly help us

Best regards

4 Replies Last Post Nov 15, 2012, 2:42 p.m. EST
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Nov 27, 2011, 2:54 p.m. EST
Hi

for me there are two level of "inverted elements".
1) your mesh is rater poor, some elements are not very regular, spiky or "thin", when you apply 2nd or higher order shape functions, the result is "inverted elements", hence COMSOL uses first order elements, but still calculates off and the reults are mostly OK (somewhat lower quality in regions with "inverted elements"
2) you work with ALE or deforming mesh, and during the solving the mesh element quality goes so low, or even the mesh truely inverts, then mostyyl te results are also wrong, and you should detect this before you get that far and do a remesh

--
Good luck
Ivar
Hi for me there are two level of "inverted elements". 1) your mesh is rater poor, some elements are not very regular, spiky or "thin", when you apply 2nd or higher order shape functions, the result is "inverted elements", hence COMSOL uses first order elements, but still calculates off and the reults are mostly OK (somewhat lower quality in regions with "inverted elements" 2) you work with ALE or deforming mesh, and during the solving the mesh element quality goes so low, or even the mesh truely inverts, then mostyyl te results are also wrong, and you should detect this before you get that far and do a remesh -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Nov 28, 2011, 8:57 a.m. EST
Hi Ivar,
Thank you for the help , I reduced the mesh size and it gave me d result. Now, I 've another problem, I 'm pretty confused with the term 'Core' in the module Electromagnetism.Can you provide some details about that??
Thanking you in advance
Yadu
Hi Ivar, Thank you for the help , I reduced the mesh size and it gave me d result. Now, I 've another problem, I 'm pretty confused with the term 'Core' in the module Electromagnetism.Can you provide some details about that?? Thanking you in advance Yadu

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Nov 15, 2012, 11:02 a.m. EST
Dear mr. Kjelberg,

I wanted to ask if you know any publications concerning the inverted mesh elements?? I need to a quote for my Bachelor theses!

Best regards,
Andreas Artelsmair
Dear mr. Kjelberg, I wanted to ask if you know any publications concerning the inverted mesh elements?? I need to a quote for my Bachelor theses! Best regards, Andreas Artelsmair

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Nov 15, 2012, 2:42 p.m. EST
Hi

No I do not really know of any articles, its references somewhere in the 4.3a doc , but its a common issue with higher order discretisation (polynomial development) on mesh elements.

Think of it: you link the node vertex of a thetraedral (4 node) mesh element shape with straight lines, then you add 2nd order polynomials, you need to consider the nabour mesh element nodes too to have enough points (minimum 3, 4 better). Then you place an extra "virtual" node/point inbetween the two thetraedral mesh nodes, but positionned on the the polynomial line. This you repeat for all (6) node pairs combinations, you are adding 6 node elements w.r.t the 4 original ones, this is a 2nd order discretization. You might also introduce a central node point at the average position from the 4 original vertex nodes.
Now if you mesh element become very slim and elongated, the intermediate node points on the mesh edges migh overlap, this means you have an inversion of the shape and the virtual volume is no longer a normal 3D topology entity, and your results will be wrong as the volume to surface rations and position of the elements are mixed up with different signes. Perhaps you can ignore the 2nd order points, you yare in linear elements and you might be able to continue to calculate, with a less denser mesh structure. But you might see that one mesh node moves "trouch the surface of the 3 others (more common perhaps in quad or square elements, then you have a fully inverted mesh elements and the results locally are wrong (sign inversion etc)

--
Good luck
Ivar
Hi No I do not really know of any articles, its references somewhere in the 4.3a doc , but its a common issue with higher order discretisation (polynomial development) on mesh elements. Think of it: you link the node vertex of a thetraedral (4 node) mesh element shape with straight lines, then you add 2nd order polynomials, you need to consider the nabour mesh element nodes too to have enough points (minimum 3, 4 better). Then you place an extra "virtual" node/point inbetween the two thetraedral mesh nodes, but positionned on the the polynomial line. This you repeat for all (6) node pairs combinations, you are adding 6 node elements w.r.t the 4 original ones, this is a 2nd order discretization. You might also introduce a central node point at the average position from the 4 original vertex nodes. Now if you mesh element become very slim and elongated, the intermediate node points on the mesh edges migh overlap, this means you have an inversion of the shape and the virtual volume is no longer a normal 3D topology entity, and your results will be wrong as the volume to surface rations and position of the elements are mixed up with different signes. Perhaps you can ignore the 2nd order points, you yare in linear elements and you might be able to continue to calculate, with a less denser mesh structure. But you might see that one mesh node moves "trouch the surface of the 3 others (more common perhaps in quad or square elements, then you have a fully inverted mesh elements and the results locally are wrong (sign inversion etc) -- Good luck Ivar

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.